Hide host effect params from docs #116670

oli-obk · 2023-10-12T16:16:05Z

addresses (only on nightly, needs backport) #116629

r? @compiler-errors

cc @GuillaumeGomez @fee1-dead

rustbot · 2023-10-12T16:16:17Z

Some changes occurred in src/librustdoc/clean/types.rs

cc @camelid

compiler-errors · 2023-10-12T16:30:15Z

src/librustdoc/clean/mod.rs

@@ -597,6 +598,7 @@ fn clean_generic_param<'tcx>(
                ty: Box::new(clean_ty(ty, cx)),
                default: default
                    .map(|ct| Box::new(ty::Const::from_anon_const(cx.tcx, ct.def_id).to_string())),
+                is_host_effect: cx.tcx.has_attr(param.def_id, sym::rustc_host),


Why are we re-encoding this here? Why not just call has_attr when needed?

this works if it comes from the HIR, but that's not always true, there is also a code path where the is_host_effect comes from a ty::GenericParamDef due to coming from a different crate. I don't have tests for this, but that's the reason why we encode this field in ty::GenericParamDef instead of just loading the information lazily.

You can call tcx.has_attr on a foreign param, though, and we should be encoding attrs for foreign params too.

I thought we were filtering out most of them.

@fee1-dead do you remember why we put the is_host_effect field on GenericParamDef? did I ask for it just to call has_attr less?

#[rustc_host] used to be encoded in an earlier version of #115727 (84a4907) but since #115727 (comment) it no longer is.

You can call tcx.has_attr on a foreign param, though, and we should be encoding attrs for foreign params too.

We currently don't encode attributes on const generics params, and IIRC I added the field for that purpose

fmease · 2023-10-12T16:35:35Z

I could be mistaken but I think this doesn't hide host arguments (I can't check out the patch rn, short on time). Consider:

#![feature(effects, const_trait_impl)]

#[const_trait]
pub trait Tr {
    fn f();
}

pub const fn g<T: ~const Tr>() { }

It's rendered as pub const fn g<T: Tr<host>, const host: bool = true>() on master.
If it's already fixed, then my bad but could you still add a test for it? Appreciate it!

tests/rustdoc/const-fn-effects.rs

fmease · 2023-10-12T16:55:55Z

I could be mistaken but I think this doesn't hide host arguments

I don't know if there're actually any ~const bounds left in core or std due to the const refactoring (again, can't check rn, commuting), so this might not be as pressing and therefore might not need to be addressed in this PR.

If you do want to handle them in this PR, then it's fine to just drop such host args outright without attempting to render ~const in rustdoc since we don't want to show that in the docs at the moment.

oli-obk · 2023-10-12T17:16:28Z

If you do want to handle them in this PR, then it's fine to just drop such host args outright without attempting to render ~const in rustdoc since we don't want to show them in the docs at the moment.

already done, just took a bit of time to find all the pieces that were needed.

compiler-errors · 2023-10-13T06:45:25Z

r? fmease since you've given the most detailed review

fmease

r=me with FIXMEs added

fmease · 2023-10-13T08:23:27Z

src/librustdoc/clean/mod.rs

@@ -597,6 +598,7 @@ fn clean_generic_param<'tcx>(
                ty: Box::new(clean_ty(ty, cx)),
                default: default
                    .map(|ct| Box::new(ty::Const::from_anon_const(cx.tcx, ct.def_id).to_string())),
+                is_host_effect: cx.tcx.has_attr(param.def_id, sym::rustc_host),


#[rustc_host] used to be encoded in an earlier version of #115727 (84a4907) but since #115727 (comment) it no longer is.

fmease · 2023-10-13T08:28:56Z

src/librustdoc/clean/utils.rs

@@ -104,6 +104,7 @@ pub(crate) fn ty_args_to_args<'tcx>(
                    arg: index,
                }),
            ))),
+            GenericArgKind::Const(ct) if let ty::ConstKind::Param(p) = ct.kind() && p.name == sym::host => None,


Contrary to the HIR code path which filters out #[rustc_host] params, this filters out params named host which isn't quite correct, right? Could you leave a FIXME(effects) if it's too annoying to fix rn?

Since we're just looking for ConstKind::Param, we obviously don't hide host args like false, true or more complex const exprs (we'd need to check the corresp. ty::Generics first I think). Could you add a FIXME(effects) for this as well / extend the previous one.

I only know of one case where this currently surfaces (without using #![feature(rustc_attrs)] + #[rustc_host]). Namely, documenting the cross-crate re-export of pub fn h<T: Tr>() {} where Tr is a const trait results in pub fn h<T>() where T: Tr<true> since we currently don't elide defaulted args in cross-crate scenarios (#80379). My PR #112463 would fix this specific case but it's blocked on performance.

(This would also surface with always-const bounds if they were introduced like T: const Trait → T: Trait<false>, etc.)

fmease · 2023-10-13T08:54:46Z

src/librustdoc/clean/mod.rs

+                    hir::GenericArg::Lifetime(_) => GenericArg::Lifetime(Lifetime::elided()),
+                    hir::GenericArg::Type(ty) => GenericArg::Type(clean_ty(ty, cx)),
+                    hir::GenericArg::Const(ct)
+                        if cx.tcx.has_attr(ct.value.def_id, sym::rustc_host) =>


This doesn't hide host args like false, true in impls. Consider:

#[const_trait] pub trait Tr {} impl Tr for () {}

It gets rendered as impl Tr<true> for () (local/HIR & cross-crate/metadata). Doesn't look super easy to fix, we'd need to look at the corresp. ty::Generics I guess? Please add a FIXME(effects).

… more const trait things in libcore

oli-obk · 2023-10-13T11:04:28Z

@bors r=fmease

bors · 2023-10-13T11:04:30Z

📌 Commit 16f8396 has been approved by fmease

It is now in the queue for this repository.

bors · 2023-10-13T12:38:21Z

⌛ Testing commit 16f8396 with merge 193e8a1...

bors · 2023-10-13T14:46:42Z

☀️ Test successful - checks-actions
Approved by: fmease
Pushing 193e8a1 to master...

a1phyr · 2023-10-17T08:17:37Z

only on nightly

The issue also exists in beta, so this needs backporting: https://doc.rust-lang.org/beta/core/ptr/struct.NonNull.html#method.dangling

GuillaumeGomez · 2023-10-17T08:49:58Z

It was already nominated to be backported to beta.

fmease · 2023-10-19T14:31:35Z

I don't know how beta-nominated rustdoc PRs get typically approved (since they're usually not showcased in the weekly t-compiler meetings afaict), typically @apiraino does it by themself looking at past PRs.

Anyway, it's no longer super important to backport this PR due to the beta-accepted #116856, right?

apiraino · 2023-10-19T14:44:59Z

correct, t-rustdoc backports are now handled by the team itself (Zulip announcement, sort of)

fmease · 2023-10-19T14:50:53Z

Oh wow, I even reacted to your Zulip message back then but I've totally forgotten about it :'D

[beta] backports and stage0 bump - Bump stage0 to released stable compiler - Hide host effect params from docs rust-lang#116670 - Fix a performance regression in obligation deduplication. rust-lang#116826 - Make `#[repr(Rust)]` and `#[repr(C)]` incompatible with one another rust-lang#116829 - Update to LLVM 17.0.3 rust-lang#116840 - Disable effects in libcore again rust-lang#116856 - revert rust-lang#114586 rust-lang#116879 r? cuviper

…ide-x-crate-host-args, r=GuillaumeGomez rustdoc: properly elide cross-crate host effect args Fixes FIXMEs introduced in rust-lang#116670.

Rollup merge of rust-lang#117531 - fmease:rustdoc-effects-properly-elide-x-crate-host-args, r=GuillaumeGomez rustdoc: properly elide cross-crate host effect args Fixes FIXMEs introduced in rust-lang#116670.

…nkov Reenable effects in libcore With rust-lang#116670, rust-lang#117531, and rust-lang#117171, I think we would be comfortable with re-enabling the effects feature for more testing in libcore. r? `@oli-obk` cc `@fmease` cc rust-lang#110395

oli-obk added 2 commits October 12, 2023 15:55

Add regression tests

20363f4

Hide host effect params from docs

c4e61fa

oli-obk added the beta-nominated Nominated for backporting to the compiler in the beta channel. label Oct 12, 2023

rustbot assigned compiler-errors Oct 12, 2023

rustbot added A-rustdoc-json Area: Rustdoc JSON backend S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-rustdoc Relevant to the rustdoc team, which will review and decide on the PR/issue. labels Oct 12, 2023

compiler-errors reviewed Oct 12, 2023

View reviewed changes

fmease reviewed Oct 12, 2023

View reviewed changes

tests/rustdoc/const-fn-effects.rs Outdated Show resolved Hide resolved

oli-obk added 2 commits October 12, 2023 16:44

Test cross crate

8f2af7e

Add regression test for generic args showing host param

cfb6afa

hide host param from generic parameter list of ~const bounds

6724f99

rustbot assigned fmease and unassigned compiler-errors Oct 13, 2023

fmease approved these changes Oct 13, 2023

View reviewed changes

Add some FIXMEs for remaining issues that we need to fix before using…

16f8396

… more const trait things in libcore

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 13, 2023

bors added the merged-by-bors This PR was explicitly merged by bors. label Oct 13, 2023

bors merged commit 193e8a1 into rust-lang:master Oct 13, 2023
11 checks passed

rustbot added this to the 1.75.0 milestone Oct 13, 2023

oli-obk deleted the host_docs branch October 17, 2023 07:03

Veykril mentioned this pull request Oct 17, 2023

Pull upstream master 2023 10 17 ferrocene/ferrocene#54

Merged

apiraino added the beta-accepted Accepted for backporting to the compiler in the beta channel. label Oct 19, 2023

cuviper mentioned this pull request Oct 21, 2023

[beta] backports and stage0 bump #117005

Merged

cuviper modified the milestones: 1.75.0, 1.74.0 Oct 21, 2023

cuviper removed the beta-nominated Nominated for backporting to the compiler in the beta channel. label Oct 21, 2023

a1phyr mentioned this pull request Oct 23, 2023

Weird <const host: bool = true> generics popping in beta docs #116629

Closed

fmease mentioned this pull request Nov 3, 2023

rustdoc: properly elide cross-crate host effect args #117531

Merged

fee1-dead mentioned this pull request Nov 12, 2023

Reenable effects in libcore #117825

Merged

fee1-dead mentioned this pull request Dec 6, 2023

Tracking Issue for removing impl const and ~const in the standard library #110395

Closed

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Hide host effect params from docs #116670

Hide host effect params from docs #116670

oli-obk commented Oct 12, 2023

rustbot commented Oct 12, 2023

compiler-errors Oct 12, 2023

oli-obk Oct 12, 2023

compiler-errors Oct 12, 2023 •

edited

Loading

oli-obk Oct 12, 2023

fmease Oct 13, 2023

fee1-dead Oct 25, 2023

fmease commented Oct 12, 2023 •

edited

Loading

fmease commented Oct 12, 2023 •

edited

Loading

oli-obk commented Oct 12, 2023

compiler-errors commented Oct 13, 2023

fmease left a comment

fmease Oct 13, 2023

fmease Oct 13, 2023 •

edited

Loading

fmease Oct 13, 2023 •

edited

Loading

fmease Oct 13, 2023 •

edited

Loading

oli-obk commented Oct 13, 2023

bors commented Oct 13, 2023

bors commented Oct 13, 2023

bors commented Oct 13, 2023

a1phyr commented Oct 17, 2023

GuillaumeGomez commented Oct 17, 2023

fmease commented Oct 19, 2023 •

edited

Loading

apiraino commented Oct 19, 2023

fmease commented Oct 19, 2023

Hide host effect params from docs #116670

Hide host effect params from docs #116670

Conversation

oli-obk commented Oct 12, 2023

rustbot commented Oct 12, 2023

compiler-errors Oct 12, 2023

Choose a reason for hiding this comment

oli-obk Oct 12, 2023

Choose a reason for hiding this comment

compiler-errors Oct 12, 2023 • edited Loading

Choose a reason for hiding this comment

oli-obk Oct 12, 2023

Choose a reason for hiding this comment

fmease Oct 13, 2023

Choose a reason for hiding this comment

fee1-dead Oct 25, 2023

Choose a reason for hiding this comment

fmease commented Oct 12, 2023 • edited Loading

fmease commented Oct 12, 2023 • edited Loading

oli-obk commented Oct 12, 2023

compiler-errors commented Oct 13, 2023

fmease left a comment

Choose a reason for hiding this comment

fmease Oct 13, 2023

Choose a reason for hiding this comment

fmease Oct 13, 2023 • edited Loading

Choose a reason for hiding this comment

fmease Oct 13, 2023 • edited Loading

Choose a reason for hiding this comment

fmease Oct 13, 2023 • edited Loading

Choose a reason for hiding this comment

oli-obk commented Oct 13, 2023

bors commented Oct 13, 2023

bors commented Oct 13, 2023

bors commented Oct 13, 2023

a1phyr commented Oct 17, 2023

GuillaumeGomez commented Oct 17, 2023

fmease commented Oct 19, 2023 • edited Loading

apiraino commented Oct 19, 2023

fmease commented Oct 19, 2023

compiler-errors Oct 12, 2023 •

edited

Loading

fmease commented Oct 12, 2023 •

edited

Loading

fmease commented Oct 12, 2023 •

edited

Loading

fmease Oct 13, 2023 •

edited

Loading

fmease Oct 13, 2023 •

edited

Loading

fmease Oct 13, 2023 •

edited

Loading

fmease commented Oct 19, 2023 •

edited

Loading